PrefixRL: Optimization of Parallel Prefix Circuits using Deep Reinforcement Learning
This work addresses the need for efficient digital circuit design, offering a novel automated method that can improve performance in high-performance computing and electronics, though it is incremental in applying RL to a specific domain.
The authors tackled the problem of designing parallel prefix circuits like adders by using deep reinforcement learning to optimize area and delay, achieving up to 30.2% lower area for the same delay compared to baselines and outperforming commercial tools in some cases.
In this work, we present a reinforcement learning (RL) based approach to designing parallel prefix circuits such as adders or priority encoders that are fundamental to high-performance digital design. Unlike prior methods, our approach designs solutions tabula rasa purely through learning with synthesis in the loop. We design a grid-based state-action representation and an RL environment for constructing legal prefix circuits. Deep Convolutional RL agents trained on this environment produce prefix adder circuits that Pareto-dominate existing baselines with up to 16.0% and 30.2% lower area for the same delay in the 32b and 64b settings respectively. We observe that agents trained with open-source synthesis tools and cell library can design adder circuits that achieve lower area and delay than commercial tool adders in an industrial cell library.